Can AI Do SEO for My Website on Its Own?

An AI tool can run most of the diagnostic work on a website faster than a person can. It can read your robots.txt, fetch pages the way a crawler does, compare what loads before and after JavaScript, and list every place your own site contradicts itself.

What it can't do is act where it has no access or no authority. It can't log into your CDN. It can't decide whether your current price is $49 or $59. It can't see a Search Console setting that has no API.

So the honest split is this:

• AI handles: detection, comparison, drafting fixes, writing the copy for a fact you confirm.
• You handle: approvals, account access, factual truth, and anything behind a login.

If you're deciding who does what across marketing and engineering, a RACI for AI visibility assigns each of these tasks to an owner.

Before deciding what to automate, it helps to know which step you're automating. Every AI answer passes through four stages.

1. Parametric. What the model already knows from training, frozen at a cutoff date. Nothing you publish this quarter changes it.
2. Access. Whether a crawler is allowed to fetch and store your page. This is set by your robots.txt, your CDN, your firewall and your rendering.
3. Retrieval. Whether your page enters the model's context for a given question. The engine decides, often after rewriting the user's question into its own search queries.
4. Generation. Which retrieved pages the model uses and cites. The model decides.

The access stage is where publishers have real control, and it's where an AI tool earns its keep. The checks are binary and the consequences are documented by the engines themselves.

Retrieval also varies by surface. As of October 2026, Google says a page must be "indexed and eligible to be shown in Google Search with a snippet" to appear as a supporting link in AI Overviews or AI Mode, with no additional technical requirements. Google renders JavaScript through Googlebot. Most other AI crawlers don't. Vercel found that ChatGPT and Claude crawlers fetch JavaScript files (11.50% and 23.84% of their requests) but don't execute them, so content that only appears after JavaScript runs may be invisible to them.

Say you run a scheduling software company with a marketing site built in a JavaScript framework. You point an AI SEO tool at it. Here's what a competent tool would flag, and who can act on each finding.

Notice the pattern. The tool finds everything in the first column. The owner closes almost every row.

The robots.txt fix is a good example of why judgement matters. OpenAI says each of its bot settings "is independent of the others." GPTBot governs training. OAI-SearchBot governs ChatGPT search. A site that wants to stay out of training but appear in ChatGPT search would write:

User-agent: GPTBot
Disallow: /

User-agent: OAI-SearchBot
Allow: /

A tool can write that file. Only you can decide whether training use is acceptable to your business.

CDN and firewall settings

robots.txt is a request, not a gate. A firewall or CDN can block an AI crawler that robots.txt allows, and the result is silent: no dashboard warns you. OpenAI's own help page says eligibility for ChatGPT search needs both an allowed OAI-SearchBot and a host or CDN that lets OpenAI's published IP addresses through.

Defaults can also change under you. From 15 September 2026, Cloudflare set Training and Agent crawlers to blocked by default on ad-displaying pages for new domains, new sites on existing accounts and all existing free-tier customers, while Search stayed allowed by default. An AI tool can detect the result. Changing it takes someone with the account login.

Accurate facts

AI platforms get specific facts wrong about specific brands. Sometimes the source is your own site, a stale price or two different founding dates. A tool can find contradictions, but it can't know which version is correct. That decision belongs to whoever owns the facts. Mistakes on your own website can cause AI errors, and the free check for contradicting facts on your own site is a quick place to start.

Stating a specific fact plainly, a price, a number or a one-line definition, on a page that can be fetched gives an engine a sentence it can quote. It's the only positive experimental signal for on-page content in the evidence.

Search Console

Google's generative AI inclusion control is opt-out: inclusion is the default for all properties. But a property can be excluded, deliberately or by inheriting the setting from a parent property, and that can't be seen from outside. Google's generative AI report also shows impressions only, with no clicks or position. Neither has an API an AI tool can read. See Google's own tools for AI search visibility for what each report shows.

Some parts of AI visibility aren't in your hands or any tool's.

• Training data. Retrieval-off answers come from what the model learned before its cutoff. Nothing you do this quarter reaches them.
• Whether retrieval fires. ChatGPT searches at the model's discretion. Google AI Overviews often don't trigger. Claude.ai search is a user toggle.
• The real queries. OpenAI says ChatGPT search rewrites your question into one or more targeted queries. You never see them.
• Which page gets cited. That's the generation stage. The model decides.

This is why any promise of a ranking, a citation or a score should make you suspicious. A critical survey of 45 GEO studies (arXiv, July 2026) found no technique with a stable, cross-platform causal effect on discoverability. A tool can remove what blocks you. It can't make a model pick you.

Run these in order. Each step depends on the one before it.

1. Check access per bot. Read robots.txt for each retrieval bot: OAI-SearchBot, PerplexityBot, Claude-SearchBot, Googlebot, bingbot. Don't confuse them with training bots like GPTBot and ClaudeBot.
2. Check the network layer. Request a page as each crawler and look for 403s or challenge pages. The AI Crawler Checker does this and shows which crawlers your firewall turns away.
3. Check rendering. Compare the raw HTML with the rendered page. Anything that matters (prices, product names, key claims) should be in the raw HTML.
4. Check suppression. Look for stray noindex, nosnippet, max-snippet:0 and data-nosnippet. Remove only the ones nobody intended.
5. Check index status engine by engine. Open Google Search Console and Bing Webmaster Tools separately. A healthy index in one tells you nothing about the other.
6. Fix your facts. Resolve contradictions, then state each key fact once, plainly, on a fetchable page.
7. Look beyond your site. When engines cite other sites instead of yours, being named on those sites is associated with appearing in the answer.

Then measure. The free checker on this site shows whether AI engines name you for up to three buyer questions. Treat it as a snapshot, and measure each engine separately rather than trusting one blended number.

Mass-publishing AI content

In OtterlyAI's 2026 experiment, two new sites with roughly 1,000 AI-generated posts each were both deindexed by Google, one after about 14 days and one after about 90, with no manual action reported. Google's AI surfaces then stopped citing them. Google's guidance says appropriate use of AI isn't against its guidelines, but its spam policy names generating many pages without adding value as scaled content abuse. The problem is volume without value, not the tool.

Following old checklists

Many AI tools still recommend tactics the evidence doesn't support. Ahrefs' matched-control study found AI Overview citations fell 4.6% after pages added schema. FAQPage schema rich results were deprecated in May 2026. Google says Search ignores llms.txt, and that you don't need to write in a special style or chunk content for AI. Existing markup does no harm. It just isn't a lever for AI answers.

Allowing GPTBot to get into ChatGPT

GPTBot governs training only. ChatGPT search uses OAI-SearchBot.

Letting the tool assert facts

If you use a model to draft pages, as covered in using ChatGPT for SEO work, check every number it writes about your company. A confident wrong price on your own site is a fact engines can repeat.

Hiding text for AI crawlers

Some tools offer hidden instructions or content shown only to AI crawlers. That's prompt injection or cloaking. It's illegitimate, and research shows it can manipulate answers, which is exactly why it's a reputational risk. Don't use it.

Will an AI SEO tool get my site cited in ChatGPT? No tool can promise that. A good tool can confirm that OAI-SearchBot is allowed, that your firewall lets it through and that your key content is in the raw HTML. Those checks make you eligible. Whether ChatGPT searches for a given question, and which page it cites, is decided by the model. Treat any guaranteed citation as a red flag.

Can I let AI write all my blog posts? You can use AI to draft, but publishing large volumes of thin pages is risky. Google's spam policy treats generating many pages without adding value for users as scaled content abuse, regardless of how they were made. In Otterly's 2026 test, both AI-only sites were removed from Google's index. Use AI as a drafting aid, then add real expertise and checked facts.

Should I block GPTBot to protect my content? That's a business decision, and it won't remove you from ChatGPT search. OpenAI documents GPTBot as its training crawler and OAI-SearchBot as its search crawler, with independent settings. You can disallow GPTBot to opt out of training and still allow OAI-SearchBot so your pages stay eligible for ChatGPT search answers.

Can an AI tool check my Search Console settings? Not the ones that matter most for AI. Google's generative AI performance report and its AI inclusion control have no API as of October 2026. The report shows impressions only. The control is on by default but can be switched off or inherited from a parent property. Someone with Search Console access has to log in and look.

Does adding schema with an AI tool help AI answers? The evidence says it isn't a lever. Google states structured data isn't required for generative AI search, and Ahrefs' matched-control study found AI Overview citations fell 4.6% on pages that added schema. Existing markup does no harm, so there's no need to remove it. Spend the time on access, rendering and accurate facts instead.